Spatial Metric Space for Pattern Recognition Problems
Authors:
Q. M. Danish Lohani,
Ashutosh Tiwari,
Mohd Shoaib Khan
Abstract:
The definition of weighted distance measure involves weights. The paper proposes a weighted distance measure without the help of weights. Here, weights are intrinsically added to the measure, and for this, the concept of metric space is generalized based on a novel divided difference operator. The proposed operator is used over a two-dimensional sequence of bounded variation, and it generalizes me…
▽ More
The definition of weighted distance measure involves weights. The paper proposes a weighted distance measure without the help of weights. Here, weights are intrinsically added to the measure, and for this, the concept of metric space is generalized based on a novel divided difference operator. The proposed operator is used over a two-dimensional sequence of bounded variation, and it generalizes metric space with the introduction of a multivalued metric space called spatial metric space. The environment considered for the study is a two-dimensional Atanassov intuitionistic fuzzy set (AIFS) under the assumption that membership and non-membership components are its independent variables. The weighted distance measure is proposed as a spatial distance measure in the spatial metric space. The spatial distance measure consists of three branches. In the first branch, there is a domination of membership values, non-membership values dominate the second branch, and the third branch is equidominant. The domination of membership and non-membership values are not in the form of weights in the proposed spatial distance measure, and hence it is a measure independent of weights. The proposed spatial metric space is mathematically studied, and as an implication, the spatial similarity measure is multivalued in nature. The spatial similarity measure can recognize a maximum of three patterns simultaneously. The spatial similarity measure is tested for the pattern recognition problems and the obtained classification results are compared with some other existing similarity measures to show its potency. This study connects the double sequence to the application domain via a divided difference operator for the first time while proposing a novel divided difference operator-based spatial metric space.
△ Less
Submitted 19 December, 2023;
originally announced December 2023.
Type-2 fuzzy reliability redundancy allocation problem and its solution using particle swarm optimization algorithm
Authors:
Zubair Ashraf,
Pranab K. Muhuri,
Q. M. Danish Lohani,
Mukul L. Roy
Abstract:
In this paper, the fuzzy multi-objective reliability redundancy allocation problem (FMORRAP) is proposed, which maximizes the system reliability while simultaneously minimizing the system cost under the type 2 fuzzy uncertainty. In the proposed formulation, the higher order uncertainties (such as parametric, manufacturing, environmental, and designers uncertainty) associated with the system are mo…
▽ More
In this paper, the fuzzy multi-objective reliability redundancy allocation problem (FMORRAP) is proposed, which maximizes the system reliability while simultaneously minimizing the system cost under the type 2 fuzzy uncertainty. In the proposed formulation, the higher order uncertainties (such as parametric, manufacturing, environmental, and designers uncertainty) associated with the system are modeled with interval type 2 fuzzy sets (IT2 FS). The footprint of uncertainty of the interval type 2 membership functions (IT2 MFs) accommodates these uncertainties by capturing the multiple opinions from several system experts. We consider IT2 MFs to represent the subsystem reliability and cost, which are to be further aggregated using extension principle to evaluate the total system reliability and cost according to their configurations, i.e., series parallel and parallel series. We proposed a particle swarm optimization (PSO) based novel solution approach to solve the FMORRAP. To demonstrate the applicability of two formulations, namely, series parallel FMORRAP and parallel series FMORRAP, we performed experimental simulations on various numerical data sets. The decision makers/system experts assign different importance to the objectives (system reliability and cost), and these preferences are represented by sets of weights. The optimal results are obtained from our solution approach, and the Pareto optimal front is established using these different weight sets. The genetic algorithm (GA) was implemented to compare the results obtained from our proposed solution approach. A statistical analysis was conducted between PSO and GA, and it was found that the PSO based Pareto solution outperforms the GA.
△ Less
Submitted 2 May, 2020;
originally announced May 2020.
